Power Estimation for VLSI Circuits Using Neural Networks

Abstract

Neural network based VLSI power estimation is done which estimates power in VLSI circuits from its input
/output and gate information, without simulation and analysis of its detail structure and the interconnections.
Artificial neural network is created which helps in estimation of power. Power estimation results from the [2] [3]
are used as the training vector for the network .The network is trained using Back-propagation algorithm. A
simple recurrent network is also introduced called Elman network which uses the back propagation for training
the network .Analysis such performance measures, regression analysis and error analysis are done to justify that
the trained network performances well. A comparative analysis on both the networks is done to show that the
neural network based approach, estimates power in faster rate. The results, concludes that the Elman network
converges faster when compared to the conventional feed forward neural networks.
Keywords: Neural Networks, VLSI, Back propagation network (BPN), Back-propagation algorithm, Recurrent
Network, Elman Neural Network (ENN)